Constant-Time Dynamic Enumeration of Word Infixes in a Regular Language

TL;DR

This paper presents a constant-time dynamic enumeration algorithm for infixes of a fixed regular language within a word, supporting updates efficiently, and characterizes the language classes where this is possible.

Contribution

It introduces a constant-time, memory-efficient algorithm for dynamic enumeration of infixes in certain regular languages, expanding understanding of language classes suitable for such algorithms.

Findings

Algorithms achieve constant delay and update time for specific language classes.

Characterizations of language classes allowing constant-memory dynamic enumeration.

Conditional lower bounds for extending results to larger language classes.

Abstract

For a fixed regular language $L$, the enumeration of $L$-infixes is the following task: we are given an input word $w = a_1 \cdots a_n$ and we must enumerate the infixes of $w$ that belong to $L$, i.e., the pairs $i \leq j$ such that $a_i \cdots a_j \in L$. We are interested in dynamic enumeration of $L$-infixes, where we must additionally support letter substitution updates on $w$ (e.g., "replace the $i$-th letter of $w$ by a letter $a$"). Each update changes the set of infixes to enumerate, and resets the enumeration state. We study for which regular languages $L$ we can perform dynamic enumeration of $L$-infixes in constant delay (i.e., the next infix is always produced in constant time) and constant additional memory throughout the enumeration, while supporting each update in constant time. We show that, for languages $L$ with a neutral letter, if the language $L$ belongs to the class ZG and is extensible (i.e., if $u \in L$ and $u$ is a factor of $v$ then $v \in L$), then dynamic enumeration of $L$-infixes can be achieved with a simple algorithm that ensures constant-time updates and constant delay, but not constant additional memory. Our main contribution is then to show an algorithm that additionally uses only constant additional memory, and applies to a more general class of semi-extensible ZG languages for which we give several equivalent characterizations. We further discuss whether our results can be generalized to larger language classes and show some (conditional) lower bounds.